(696d) A Novel MILP Model for Maintenance Scheduling of Gas Turbines

All industrial sites require regular maintenance to enhance reliability of their equipments and avoid emergency shutdowns. The main concerns of maintenance scheduling are to guarantee feasible material and utility balances while minimizing payment for skilled labor [1]. As in-house skilled labors are limited and external labor is expensive, in order to minimize salary expenditure, plants will maintain separately to reduce the manpower requirement at a time.

Using information from a long-term maintenance plant as input, Cheung et al. [1] used a discrete-time MILP formulation to schedule the maintenance of 7 plants over a period of one month. The maintenance demand profile included shutdown and startup periods of fixed length and a repair period of variable length (above a minimum time) to provide additional degrees of freedom for increasing profit.

Jain & Grossmann [2] proposed a MINLP continuous-time scheduling model for maintenance planning of continuous multiproduct plants. The model is more detailed than the one in [1] in the sense that the performance of the plant is assumed to decrease exponentially with time (measured after previous cleaning). In the context of a cyclic schedule, the outputs are the number of cycles, processing time per product and the cycle time, being the objective the maximization of profit. The cycle time for the different problems went from one to a few months.

In this work, we consider the maintenance scheduling problem of a power plant comprising N natural gas engines for electricity production. Minimum electricity production requires N-8 engines operating simultaneously, with the remainder either on standby or shutdown mode. Furthermore, there is a single maintenance team doing shutdowns, meaning that only one engine can be on shutdown at a given time. The maintenance schedule of a particular gas engine is based on run hours, not considering the hours spent in standby mode, and includes shutdowns every H hours (some flexibility is allowed). These have a fixed duration that is a function of the number of previous shutdowns. The maintenance schedule is to be derived for a period of several years, featuring multiple shutdowns per engine.

We propose a novel mixed-integer linear programming formulation that handles the resource constraints and the time dependent degradation of the turbines. It relies on a hybrid continuous-time formulation featuring multiple time grids [3] to keep track of the events taking place at the gas engines, and sequencing variables to enforce the single maintenance crew constraint. Conceptually similar models have been proposed for automated wet-etching stations involving a single robot for material transfer [4-5].

References:

[1]- Cheung K-Y, Hui C-W, Sakamoto H, Hirata K, O’Young L. Short-term site-wide maintenance scheduling. Comp. Chem. Eng. 2004; 28: 91-102.

[2] Jain V, Grossmann I.E. Scheduling of Continuous Parallel-Process Units with Decaying Performance. AIChE J. 1998; 44(7): 1623-1636.

[3]- Castro PM, Grossmann IE. New Continuous-time MILP model for the short-term scheduling of multistage batch plants. Ind. Eng. Chem. Res. 2005; 44: 9175-9190.

[4] Bhushan S, Karimi IA. An MILP Approach to Automated Wet-Etch Station Scheduling. Ind. Eng. Chem. Res. 2003; 42: 1391-1399.

[5] Castro PM, Zeballos LJ, Méndez CA. Hybrid Time Slots Sequencing Model for a Class of Scheduling Problems. AIChE J. 2012; 58(3): 789-800.